Molecular mechanisms of gene silencing mediated by DNA methylation

Abstract

DNA methylation and chromatin modification operate along a common pathway to repress transcription; accordingly, several experiments demonstrate that the effects of DNA methylation can spread in cis and do not require promoter modification. In order to investigate the molecular details of the inhibitory effect of methylation, we microinjected into Xenopus oocytes a series of constructs containing a human CpG-rich sequence which has been differentially methylated and cloned at different positions relative to a specific promoter. The parameters influencing the diffusion of gene silencing and the importance of histone deacetylation in the spreading effect were analyzed. We demonstrate that a few methylated cytosines can inhibit a flanking promoter but a threshold of modified sites is required to organize a stable, diffusible chromatin structure. Histone deacetylation is the main cause of gene repression only when methylation does not reach levels sufficient to establish this particular structure. Moreover, contrary to the common thought, promoter modification does not lead to the greater repressive effect; the existence of a competition between transactivators and methyl-binding proteins for the establishment of an open conformation justifies the results obtained.

Figures

FIG. 1.

Schematic representation of the experimental strategy. (A) The 241-bp sequence amplified from the human aprt CpG island is shown. Methylatable CpG dinucleotides are indicated in bold letters. (B) Scheme of the protocol for the generation of differentially methylated templates. (C) Evaluation of the correct pattern of methylation. The methylation pattern of 11-aprt-tkCAT was analyzed by digestion with the methylation-sensitive restriction enzyme HpaII. The digestion profile was directly visualized on a 1% agarose gel (lanes 5 and 6), or the restricted DNA amplified by PCR (lanes 10 to 13). Markers of low-mass DNA (lane 1), λ BstEII-cut DNA (lanes 2), λ HindIII/φχ HaeIII-cut DNA (lane 7), and 100-bp DNA ladder (lane 8) were resolved. The positions of the supercoiled gel-purified mock-methylated and methylated DNAs are shown in lanes 3 and 4, respectively. Lane 5, HpaII-digested mock-methylated DNA. Lane 6, HpaII-digested regionally methylated DNA. The arrow in lane 6 indicates the appearance of a longer DNA fragment indicative of the incapacity of the enzyme to digest the methylated aprt fragment. PCR amplification of the HpaII-restricted DNAs shown in each lane (10 to 13), and the presence of the internal control is indicated with an asterisk. In contrast, the 450-bp fragment, obtainable only if the aprt fragment is methylated, appears only when the digested regionally methylated DNA is used for the amplification (lanes 10 and 11) and not when mock-methylated DNA is analyzed (lanes 12 and 13). In lane 9, a no-DNA control was loaded.

FIG. 2.

A short region of methylated DNA severely affects tk expression. (A) Schematic representation of the regionally methylated 14-aprt-tkCAT and 11-aprt-tkCAT plasmids. The methylated sites are depicted by lollipops. (B) Methylated (CH3; lanes 2, 4, 6, and 8) or mock-methylated (mock; lanes 1, 3, 5, and 7) plasmids 14-aprt-tkCAT (lanes 1, 2, 5, and 6) and 11-aprt-tkCAT (lanes 3, 4, 7, and 8) were injected into oocyte nuclei, and transcription from the HSV tk promoter was assayed by primer extension at 2 h (lanes 1 to 4) and 16 h (lanes 5 to 8) after injection. Coinjection of pCMVCAT served as an internal standard. The positions of correctly initiated transcripts from the tk promoter (tk) and the CMV promoter (CMV) are indicated. (C) Total DNA was isolated from oocytes 16 h after injection. To evaluate if comparable amounts of mock-methylated (lanes 1 and 3) or patch-methylated (lanes 2 and 4) DNAs were injected, purified DNA was analyzed by Southern blot using the tk promoter fragment as a probe. Positions of supercoiled (sc), linear (lin), and relaxed (rel) DNA are indicated. (D) Methylated or mock-methylated 14-aprt-tkCAT or 11-aprt-tkCAT DNAs were injected together with the CMV internal standard, and primer extension was performed 16 h later. Transcription signals were quantitated with a PhosphorImager, and tkCAT transcription was standardized against the internal standard. tkCAT inhibition is expressed as a percentage of the activity of the 100% active mock-methylated control and is indicated as the mean value calculated from three independent experiments. Vertical bars represent standard deviations.

FIG. 3.

Ability of a methylated DNA to influence an adjacent promoter is a function of the number of modified sites. The 11-aprt-tkCAT, 6-aprt-tkCAT, 4-aprt-tkCAT, 3-aprt-tkCAT, and 1-aprt-tkCAT templates (see Materials and Methods) were microinjected and transcription(A) and methylation stability (B) were assayed 16 h after injection. (A) Transcription signals obtained from primer extensions were quantitated with a PhosphorImager, and tk inhibition, expressed as a percentage of repression related to the mock-methylated control, is reported. The graph plots the means of triplicate determinations, and the error bars show the standard deviation. (B) Total DNA was isolated from injected oocytes, and the methylation pattern was analyzed by digestion with the methylation-sensitive restriction enzyme HpaII and subsequent Southern blot analysis. A fragment spanning from the aprt sequence to the end of the CAT gene was used as the probe. The fragments expected from the digestion of the mock-methylated and methylated templates are indicated in the scheme. In bold are predicted fragments expected only if the DNAs are correctly methylated.

FIG. 4.

Diffusion in cis of the repressive effect is influenced by the number of methylated sites. (A) Repressive activity of the regionally methylated 11-aprt-tkCAT template was compared to that of patch-methylated 11-aprt-tkCAT+100 or 11-aprt-tkCAT+700 or the differentially modified tkCAT-11aprt. Transcription was assayed 16 h after injection, quantified, and standardized against the CMV control compared to the corresponding mock controls. (B) As in panel A, but the inhibitory ability of the modified 6-aprt-tkCAT was compared to that of 6-aprt-tkCAT+100, 6-aprt-tkCAT+700, and tkCAT-6aprt. (C) As in panel A but using the 4-aprt-tkCAT template and the 4-aprt-tkCAT+100 vector. (D) As in panel A but using the 3-aprt-tkCAT DNA and 3-aprt-tkCAT+100. The error bars represent the standard deviations of three independent experiments.

FIG. 5.

Trichostatin A relieves transcriptional inhibition only when a limited number of CpGs are modified. Xenopus oocytes were injected with mock-methylated or regionally methylated DNAs (11-aprt-tkCAT, 6-aprt-tkCAT, 4-aprt-tkCAT, and tkCAT-11aprt) and incubated in the absence (−) or presence (+) of trichostatin A. (A) Transcription was evaluated by primer extension 16 h after injection. Transcriptional signals were quantified with a PhosphorImager, and the levels of transcription relative to DNA mass injected (estimated from Southern blotting, panel B) are indicated in the graph. Data represent the mean values obtained from three independent experiments; vertical bars represent standard deviations. (B) DNA was extracted from half of the collected oocytes and analyzed by electrophoresis on a 1% agarose gel before transfer to a membrane and hybridization with an aprt-tk-CAT probe. Lanes 1 and 2 indicate positions of supercoiled (sc) and linear (lin) DNA. In lanes 3 to 18, purified DNAs were loaded following the same order as in panel A.

FIG. 6.

Transcriptional activators compete with methylation-specific repressors in vivo. (A) M13-tkCAT vector methylated only on the promoter (lane 2) or the corresponding unmethylated DNA (lane 1) was microinjected. Transcription was evaluated by primer extension and quantitated with a PhosphorImager. Transcriptional levels, relative to the CMV internal standard, were calculated and are reported in the graph as percentages of the 100% active mock DNA. The mean value and standard deviation of three independent experiments are reported. In the upper part of the panel, the regionally methylated M13-tkCAT DNA, characterized by 23 mCpGs distributed over the HSV tk sequence, is represented. (B) TBP (lane 3) or GAL4VP16 (lane 4) mRNA was injected into stage VI oocytes. After incubation for 16 h, a mixture of templates containing unmodified Gal4×5-tkCAT DNA (lane 1) or the same regionally methylated template (lanes 2 to 4) and the CMVCAT internal control was microinjected. Oocytes were collected 16 h after DNA injection, and transcription was analyzed. In the upper part, as above, is a scheme of the partially methylated Gal4×5-tkCAT DNA. (C) A total of 6 ng of TBP (lane 1) or GAL4VP16 (lane 3) mRNAs was injected, and oocytes were incubated for 16 h in the presence of [35S]methionine (100 μCi/ml). Batches of 10 oocytes were collected, and a whole-cell extract (1 oocyte equivalent) from uninjected (lanes 2 and 4) or injected (lanes 1 and 3) oocytes was loaded on SDS-10% PAGE followed by fluorography. The correct positions of exogenously expressed TBP and GAL4VP16 proteins are indicated. (D) A total of 6 ng of a specific mRNA (GAL4, GAL4VP16, a mixture of GALTR and RXR, GAL4OCT1, GAL4OCT2, and TBP) was injected, oocytes were incubated for 16 h and subsequently reinjected with a mixture of regionally methylated Gal4×5-tkCAT and CMV DNAs. A mixture of unmethylated Gal4×5-tkCAT and CMV DNA was also injected as a control (mock). Thyroid hormone (T3, 50 nM) was added to the oocytes synthesizing GALTR/RXR, and all samples were incubated overnight. Transcription was analyzed by primer extension and quantified with a PhosphorImager. The histogram reports the percentage of transcription relative to CMV, calculated from two independent experiments.

FIG. 7.

Diffusion of methylation-dependent silencing is blocked by exogenous GAL4VP16 but not TBP. GAL4 (lane 3), GAL4VP16 (lane 4), or TBP (lane 5) mRNA was injected into stage VI oocytes. After 16 h, a mixture of templates containing the regionally methylated 11-aprtGal4×5-tkCAT and CMVCAT DNAs was injected (lanes 2 to 5). In lane 1, the injected DNA mixture contained the mock-methylated vector together with CMVCAT. Oocytes were incubated for 16 h, and RNA was extracted and analyzed by primer extension. The upper part of the figure shows a scheme of the regionally methylated 11-aprtGal4×5-tkCAT template.

FIG. 8.

Model for the molecular mechanisms of gene silencing mediated by DNA methylation. The left side of the model represents the repressive mechanism determined by a small number of modified dinucleotides; on the right, the hypothetical mechanism by which a large number of methylated dinucleotides exert a long-distance inhibitory effect is reported. In this cartoon, DNA is depicted as a black line wrapped around nucleosomes (green), methylcytosines are indicated with red circles, histone tails with a chain of black circles, methyl-binding proteins with yellow triangles, histone deacetylases (HDAC) with blue rectangles, and remodeling factors with orange ovals.